ABSTRACT
OBJECTIVE: Evidence suggests that the most promising results in interictal localization of the epileptogenic zone (EZ) are achieved by a combination of multiple stereo-electroencephalography (SEEG) biomarkers in machine learning models. These biomarkers usually include SEEG features calculated in standard frequency bands, but also high-frequency (HF) bands. Unfortunately, HF features require extra effort to record, store, and process. Here we investigate the added value of these HF features for EZ localization and postsurgical outcome prediction. METHODS: In 50 patients we analyzed 30 min of SEEG recorded during non-rapid eye movement sleep and tested a logistic regression model with three different sets of features. The first model used broadband features (1-500 Hz); the second model used low-frequency features up to 45 Hz; and the third model used HF features above 65 Hz. The EZ localization by each model was evaluated by various metrics including the area under the precision-recall curve (AUPRC) and the positive predictive value (PPV). The differences between the models were tested by the Wilcoxon signed-rank tests and Cliff's Delta effect size. The differences in outcome predictions based on PPV values were further tested by the McNemar test. RESULTS: The AUPRC score of the random chance classifier was .098. The models (broad-band, low-frequency, high-frequency) achieved median AUPRCs of .608, .582, and .522, respectively, and correctly predicted outcomes in 38, 38, and 33 patients. There were no statistically significant differences in AUPRC or any other metric between the three models. Adding HF features to the model did not have any additional contribution. SIGNIFICANCE: Low-frequency features are sufficient for correct localization of the EZ and outcome prediction with no additional value when considering HF features. This finding allows significant simplification of the feature calculation process and opens the possibility of using these models in SEEG recordings with lower sampling rates, as commonly performed in clinical routines.
ABSTRACT
OBJECTIVE: Interictal biomarkers of the epileptogenic zone (EZ) and their use in machine learning models open promising avenues for improvement of epilepsy surgery evaluation. Currently, most studies restrict their analysis to short segments of intracranial EEG (iEEG). METHODS: We used 2381 hours of iEEG data from 25 patients to systematically select 5-minute segments across various interictal conditions. Then, we tested machine learning models for EZ localization using iEEG features calculated within these individual segments or across them and evaluated the performance by the area under the precision-recall curve (PRAUC). RESULTS: On average, models achieved a score of 0.421 (the result of the chance classifier was 0.062). However, the PRAUC varied significantly across the segments (0.323-0.493). Overall, NREM sleep achieved the highest scores, with the best results of 0.493 in N2. When using data from all segments, the model performed significantly better than single segments, except NREM sleep segments. CONCLUSIONS: The model based on a short segment of iEEG recording can achieve similar results as a model based on prolonged recordings. The analyzed segment should, however, be carefully and systematically selected, preferably from NREM sleep. SIGNIFICANCE: Random selection of short iEEG segments may give rise to inaccurate localization of the EZ.
Subject(s)
Electroencephalography , Epilepsy , Machine Learning , Humans , Female , Male , Adult , Epilepsy/physiopathology , Epilepsy/diagnosis , Electroencephalography/methods , Middle Aged , Time Factors , Young Adult , Electrocorticography/methods , Electrocorticography/standards , Adolescent , Brain/physiopathology , Sleep Stages/physiologyABSTRACT
Objective.The current practices of designing neural networks rely heavily on subjective judgment and heuristic steps, often dictated by the level of expertise possessed by architecture designers. To alleviate these challenges and streamline the design process, we propose an automatic method, a novel approach to enhance the optimization of neural network architectures for processing intracranial electroencephalogram (iEEG) data.Approach.We present a genetic algorithm, which optimizes neural network architecture and signal pre-processing parameters for iEEG classification.Main results.Our method improved the macroF1 score of the state-of-the-art model in two independent datasets, from St. Anne's University Hospital (Brno, Czech Republic) and Mayo Clinic (Rochester, MN, USA), from 0.9076 to 0.9673 and from 0.9222 to 0.9400 respectively.Significance.By incorporating principles of evolutionary optimization, our approach reduces the reliance on human intuition and empirical guesswork in architecture design, thus promoting more efficient and effective neural network models. The proposed method achieved significantly improved results when compared to the state-of-the-art benchmark model (McNemar's test,p⪠0.01). The results indicate that neural network architectures designed through machine-based optimization outperform those crafted using the subjective heuristic approach of a human expert. Furthermore, we show that well-designed data preprocessing significantly affects the models' performance.
Subject(s)
Electrocorticography , Neural Networks, Computer , Humans , Electroencephalography/methods , Signal Processing, Computer-AssistedABSTRACT
Manual visual review, annotation and categorization of electroencephalography (EEG) is a time-consuming task that is often associated with human bias and requires trained electrophysiology experts with specific domain knowledge. This challenge is now compounded by development of measurement technologies and devices allowing large-scale heterogeneous, multi-channel recordings spanning multiple brain regions over days, weeks. Currently, supervised deep-learning techniques were shown to be an effective tool for analyzing big data sets, including EEG. However, the most significant caveat in training the supervised deep-learning models in a clinical research setting is the lack of adequate gold-standard annotations created by electrophysiology experts. Here, we propose a semi-supervised machine learning technique that utilizes deep-learning methods with a minimal amount of gold-standard labels. The method utilizes a temporal autoencoder for dimensionality reduction and a small number of the expert-provided gold-standard labels used for kernel density estimating (KDE) maps. We used data from electrophysiological intracranial EEG (iEEG) recordings acquired in two hospitals with different recording systems across 39 patients to validate the method. The method achieved iEEG classification (Pathologic vs. Normal vs. Artifacts) results with an area under the receiver operating characteristic (AUROC) scores of 0.862 ± 0.037, 0.879 ± 0.042, and area under the precision-recall curve (AUPRC) scores of 0.740 ± 0.740, 0.714 ± 0.042. This demonstrates that semi-supervised methods can provide acceptable results while requiring only 100 gold-standard data samples in each classification category. Subsequently, we deployed the technique to 12 novel patients in a pseudo-prospective framework for detecting Interictal epileptiform discharges (IEDs). We show that the proposed temporal autoencoder was able to generalize to novel patients while achieving AUROC of 0.877 ± 0.067 and AUPRC of 0.705 ± 0.154.
Subject(s)
Electrocorticography , Electroencephalography , Humans , Prospective Studies , Electroencephalography/methods , Brain/physiology , ROC CurveABSTRACT
While various QRS detection and classification methods were developed in the past, the Holter ECG data acquired during daily activities by wearable devices represent new challenges such as increased noise and artefacts due to patient movements. Here, we present a deep-learning model to detect and classify QRS complexes in single-lead Holter ECG. We introduce a novel approach, delivering QRS detection and classification in one inference step. We used a private dataset (12,111 Holter ECG recordings, length of 30 s) for training, validation, and testing the method. Twelve public databases were used to further test method performance. We built a software tool to rapidly annotate QRS complexes in a private dataset, and we annotated 619,681 QRS complexes. The standardised and down-sampled ECG signal forms a 30-s long input for the deep-learning model. The model consists of five ResNet blocks and a gated recurrent unit layer. The model's output is a 30-s long 4-channel probability vector (no-QRS, normal QRS, premature ventricular contraction, premature atrial contraction). Output probabilities are post-processed to receive predicted QRS annotation marks. For the QRS detection task, the proposed method achieved the F1 score of 0.99 on the private test set. An overall mean F1 cross-database score through twelve external public databases was 0.96 ± 0.06. In terms of QRS classification, the presented method showed micro and macro F1 scores of 0.96 and 0.74 on the private test set, respectively. Cross-database results using four external public datasets showed micro and macro F1 scores of 0.95 ± 0.03 and 0.73 ± 0.06, respectively. Presented results showed that QRS detection and classification could be reliably computed in one inference step. The cross-database tests showed higher overall QRS detection performance than any of compared methods.
Subject(s)
Ventricular Premature Complexes , Wearable Electronic Devices , Algorithms , Artifacts , Electrocardiography/methods , Electrocardiography, Ambulatory/methods , Humans , Signal Processing, Computer-AssistedABSTRACT
BACKGROUND: Treating memory and cognitive deficits requires knowledge about anatomical sites and neural activities to be targeted with particular therapies. Emerging technologies for local brain stimulation offer attractive therapeutic options but need to be applied to target specific neural activities, at distinct times, and in specific brain regions that are critical for memory formation. METHODS: The areas that are critical for successful encoding of verbal memory as well as the underlying neural activities were determined directly in the human brain with intracranial electrophysiological recordings in epilepsy patients. We recorded a broad range of spectral activities across the cortex of 135 patients as they memorised word lists for subsequent free recall. FINDINGS: The greatest differences in the spectral power between encoding subsequently recalled and forgotten words were found in low theta frequency (3-5 Hz) activities of the left anterior prefrontal cortex. This subsequent memory effect was proportionally greater in the lower frequency bands and in the more anterior cortical regions. We found the peak of this memory signal in a distinct part of the prefrontal cortex at the junction between the Broca's area and the frontal pole. The memory effect in this confined area was significantly higher (Tukey-Kramer test, p<0.05) than in other anatomically distinct areas. INTERPRETATION: Our results suggest a focal hotspot of human verbal memory encoding located in the higher-order processing region of the prefrontal cortex, which presents a prospective target for modulating cognitive functions in the human patients. The memory effect provides an electrophysiological biomarker of low frequency neural activities, at distinct times of memory encoding, and in one hotspot location in the human brain. FUNDING: Open-access datasets were originally collected as part of a BRAIN Initiative project called Restoring Active Memory (RAM) funded by the Defence Advanced Research Project Agency (DARPA). CT, ML, MTK and this research were supported from the First Team grant of the Foundation for Polish Science co-financed by the European Union under the European Regional Development Fund.
Subject(s)
Memory , Prefrontal Cortex , Brain/physiology , Brain Mapping , Humans , Magnetic Resonance Imaging , Memory/physiology , Mental Recall/physiology , Prefrontal Cortex/physiologyABSTRACT
Early implantable epilepsy therapy devices provided open-loop electrical stimulation without brain sensing, computing, or an interface for synchronized behavioural inputs from patients. Recent epilepsy stimulation devices provide brain sensing but have not yet developed analytics for accurately tracking and quantifying behaviour and seizures. Here we describe a distributed brain co-processor providing an intuitive bi-directional interface between patient, implanted neural stimulation and sensing device, and local and distributed computing resources. Automated analysis of continuous streaming electrophysiology is synchronized with patient reports using a handheld device and integrated with distributed cloud computing resources for quantifying seizures, interictal epileptiform spikes and patient symptoms during therapeutic electrical brain stimulation. The classification algorithms for interictal epileptiform spikes and seizures were developed and parameterized using long-term ambulatory data from nine humans and eight canines with epilepsy, and then implemented prospectively in out-of-sample testing in two pet canines and four humans with drug-resistant epilepsy living in their natural environments. Accurate seizure diaries are needed as the primary clinical outcome measure of epilepsy therapy and to guide brain-stimulation optimization. The brain co-processor system described here enables tracking interictal epileptiform spikes, seizures and correlation with patient behavioural reports. In the future, correlation of spikes and seizures with behaviour will allow more detailed investigation of the clinical impact of spikes and seizures on patients.
ABSTRACT
Objective. This paper introduces a winning solution (team ISIBrno-AIMT) to the official round of PhysioNet Challenge 2021. The main goal of the challenge was a classification of ECG recordings into 26 multi-label pathological classes with a variable number of leads (e.g. 12, 6, 4, 3, 2). The main objective of this study is to verify whether the multi-head-attention mechanism influences the model performance.Approach. We introduced an ECG classification method based on the ResNet architecture with a multi-head attention mechanism for the official round of the challenge. However, empirical findings collected during model development suggested that the multi-head attention layer might not significantly impact the final classification performance. For this reason, during the follow-up round, we removed a multi-head attention layer to test the influence on model performance. Like the official round, the model is optimized using a mixture of loss functions, i.e. binary cross-entropy, custom challenge score loss function, and custom sparsity loss function. Probability thresholds for each classification class are estimated using the evolutionary optimization method. The final architecture consists of three submodels forming a majority voting classification ensemble.Main results. The modified model without the multi-head attention layer increased the overall challenge score to 0.59 compared to the 0.58 from the official round.Significance. Our findings from the follow-up submission support the fact that the multi-head attention layer in the proposed architecture does not significantly affect the classification performance.
Subject(s)
Algorithms , Electrocardiography , Electrocardiography/methods , Entropy , ProbabilityABSTRACT
Objective.Electrical deep brain stimulation (DBS) is an established treatment for patients with drug-resistant epilepsy. Sleep disorders are common in people with epilepsy, and DBS may actually further disturb normal sleep patterns and sleep quality. Novel implantable devices capable of DBS and streaming of continuous intracranial electroencephalography (iEEG) signals enable detailed assessments of therapy efficacy and tracking of sleep related comorbidities. Here, we investigate the feasibility of automated sleep classification using continuous iEEG data recorded from Papez's circuit in four patients with drug resistant mesial temporal lobe epilepsy using an investigational implantable sensing and stimulation device with electrodes implanted in bilateral hippocampus (HPC) and anterior nucleus of thalamus (ANT).Approach.The iEEG recorded from HPC is used to classify sleep during concurrent DBS targeting ANT. Simultaneous polysomnography (PSG) and sensing from HPC were used to train, validate and test an automated classifier for a range of ANT DBS frequencies: no stimulation, 2 Hz, 7 Hz, and high frequency (>100 Hz).Main results.We show that it is possible to build a patient specific automated sleep staging classifier using power in band features extracted from one HPC iEEG sensing channel. The patient specific classifiers performed well under all thalamic DBS frequencies with an average F1-score 0.894, and provided viable classification into awake and major sleep categories, rapid eye movement (REM) and non-REM. We retrospectively analyzed classification performance with gold-standard PSG annotations, and then prospectively deployed the classifier on chronic continuous iEEG data spanning multiple months to characterize sleep patterns in ambulatory patients living in their home environment.Significance.The ability to continuously track behavioral state and fully characterize sleep should prove useful for optimizing DBS for epilepsy and associated sleep, cognitive and mood comorbidities.
Subject(s)
Anterior Thalamic Nuclei , Deep Brain Stimulation , Sleep Wake Disorders , Brain , Deep Brain Stimulation/methods , Epilepsy/complications , Hippocampus , Humans , Retrospective Studies , Sleep Wake Disorders/complications , Sleep Wake Disorders/diagnosis , Sleep Wake Disorders/therapy , ThalamusABSTRACT
Chronic brain recordings suggest that seizure risk is not uniform, but rather varies systematically relative to daily (circadian) and multiday (multidien) cycles. Here, one human and seven dogs with naturally occurring epilepsy had continuous intracranial EEG (median 298 days) using novel implantable sensing and stimulation devices. Two pet dogs and the human subject received concurrent thalamic deep brain stimulation (DBS) over multiple months. All subjects had circadian and multiday cycles in the rate of interictal epileptiform spikes (IES). There was seizure phase locking to circadian and multiday IES cycles in five and seven out of eight subjects, respectively. Thalamic DBS modified circadian (all 3 subjects) and multiday (analysis limited to the human participant) IES cycles. DBS modified seizure clustering and circadian phase locking in the human subject. Multiscale cycles in brain excitability and seizure risk are features of human and canine epilepsy and are modifiable by thalamic DBS.
Subject(s)
Deep Brain Stimulation/methods , Epilepsy/prevention & control , Seizures/prevention & control , Thalamus/physiology , Animals , Circadian Rhythm , Dogs , Electroencephalography , Humans , RiskABSTRACT
A wide spectrum of brain rhythms are engaged throughout the human cortex in cognitive functions. How the rhythms of various frequency ranges are coordinated across the space of the human cortex and time of memory processing is inconclusive. They can either be coordinated together across the frequency spectrum at the same cortical site and time or induced independently in particular bands. We used a large dataset of human intracranial electroencephalography (iEEG) to parse the spatiotemporal dynamics of spectral activities induced during formation of verbal memories. Encoding of words for subsequent free recall activated low frequency theta, intermediate frequency alpha and beta, and high frequency gamma power in a mosaic pattern of discrete cortical sites. A majority of the cortical sites recorded activity in only one of these frequencies, except for the visual cortex where spectral power was induced across multiple bands. Each frequency band showed characteristic dynamics of the induced power specific to cortical area and hemisphere. The power of the low, intermediate, and high frequency activities propagated in independent sequences across the visual, temporal and prefrontal cortical areas throughout subsequent phases of memory encoding. Our results provide a holistic, simplified model of the spectral activities engaged in the formation of human memory, suggesting an anatomically and temporally distributed mosaic of coordinated brain rhythms.
Subject(s)
Electroencephalography/methods , Memory/physiology , Adult , Datasets as Topic , Epilepsy/diagnostic imaging , Epilepsy/surgery , Female , Humans , Magnetic Resonance Imaging , Male , Tomography, X-Ray ComputedABSTRACT
Epilepsy is one of the most common neurological disorders, and it affects almost 1% of the population worldwide. Many people living with epilepsy continue to have seizures despite anti-epileptic medication therapy, surgical treatments, and neuromodulation therapy. The unpredictability of seizures is one of the most disabling aspects of epilepsy. Furthermore, epilepsy is associated with sleep, cognitive, and psychiatric comorbidities, which significantly impact the quality of life. Seizure predictions could potentially be used to adjust neuromodulation therapy to prevent the onset of a seizure and empower patients to avoid sensitive activities during high-risk periods. Long-term objective data is needed to provide a clearer view of brain electrical activity and an objective measure of the efficacy of therapeutic measures for optimal epilepsy care. While neuromodulation devices offer the potential for acquiring long-term data, available devices provide very little information regarding brain activity and therapy effectiveness. Also, seizure diaries kept by patients or caregivers are subjective and have been shown to be unreliable, in particular for patients with memory-impairing seizures. This paper describes the design, architecture, and development of the Mayo Epilepsy Personal Assistant Device (EPAD). The EPAD has bi-directional connectivity to the implanted investigational Medtronic Summit RC+STM device to implement intracranial EEG and physiological monitoring, processing, and control of the overall system and wearable devices streaming physiological time-series signals. In order to mitigate risk and comply with regulatory requirements, we developed a Quality Management System (QMS) to define the development process of the EPAD system, including Risk Analysis, Verification, Validation, and protocol mitigations. Extensive verification and validation testing were performed on thirteen canines and benchtop systems. The system is now under a first-in-human trial as part of the US FDA Investigational Device Exemption given in 2018 to study modulated responsive and predictive stimulation using the Mayo EPAD system and investigational Medtronic Summit RC+STM in ten patients with non-resectable dominant or bilateral mesial temporal lobe epilepsy. The EPAD system coupled with an implanted device capable of EEG telemetry represents a next-generation solution to optimizing neuromodulation therapy.
ABSTRACT
Intracranial electroencephalographic (iEEG) recordings from patients with epilepsy provide distinct opportunities and novel data for the study of co-occurring psychiatric disorders. Comorbid psychiatric disorders are very common in drug-resistant epilepsy and their added complexity warrants careful consideration. In this review, we first discuss psychiatric comorbidities and symptoms in patients with epilepsy. We describe how epilepsy can potentially impact patient presentation and how these factors can be addressed in the experimental designs of studies focused on the electrophysiologic correlates of mood. Second, we review emerging technologies to integrate long-term iEEG recording with dense behavioral tracking in naturalistic environments. Third, we explore questions on how best to address the intersection between epilepsy and psychiatric comorbidities. Advances in ambulatory iEEG and long-term behavioral monitoring technologies will be instrumental in studying the intersection of seizures, epilepsy, psychiatric comorbidities, and their underlying circuitry.
ABSTRACT
BACKGROUND: While the effects of prolonged sleep deprivation (≥24 h) on seizure occurrence has been thoroughly explored, little is known about the effects of day-to-day variations in the duration and quality of sleep on seizure probability. A better understanding of the interaction between sleep and seizures may help to improve seizure management. METHODS: To explore how sleep and epileptic seizures are associated, we analysed continuous intracranial electroencephalography (EEG) recordings collected from 10 patients with refractory focal epilepsy undergoing ordinary life activities between 2010 and 2012 from three clinical centres (Austin Health, The Royal Melbourne Hospital, and St Vincent's Hospital of the Melbourne University Epilepsy Group). A total of 4340 days of sleep-wake data were analysed (average 434 days per patient). EEG data were sleep scored using a semi-automated machine learning approach into wake, stages one, two, and three non-rapid eye movement sleep, and rapid eye movement sleep categories. FINDINGS: Seizure probability changes with day-to-day variations in sleep duration. Logistic regression models revealed that an increase in sleep duration, by 1·66 ± 0·52 h, lowered the odds of seizure by 27% in the following 48 h. Following a seizure, patients slept for longer durations and if a seizure occurred during sleep, then sleep quality was also reduced with increased time spent aroused from sleep and reduced rapid eye movement sleep. INTERPRETATION: Our results suggest that day-to-day deviations from regular sleep duration correlates with changes in seizure probability. Sleeping longer, by 1·66 ± 0·52 h, may offer protective effects for patients with refractory focal epilepsy, reducing seizure risk. Furthermore, the occurrence of a seizure may disrupt sleep patterns by elongating sleep and, if the seizure occurs during sleep, reducing its quality.
ABSTRACT
OBJECTIVE: Our goal was to analyze the electrophysiological response to direct electrical stimulation (DES) systematically applied at a wide range of parameters and anatomical sites, with particular focus on neural activities associated with memory and cognition. METHODS: We used a large set of intracranial EEG (iEEG) recordings with DES from 45 subjects with electrodes implanted both subdurally on the cortical surface and subcortically into the brain parenchyma. Subjects were stimulated in blocks of alternating frequency and amplitude parameters during quiet wakefulness. RESULTS: Stimulating at different frequencies and amplitudes of electric current revealed a persistent pattern of response in the slow and the fast neural activities. In particular, amplification of the theta (4-7 Hz) and attenuation of the gamma (29-52 Hz) power-in-band was observed with increasing the stimulation parameters. This opposite effect on the low and high frequency bands was found across a network of selected local and distal sites proportionally to the proximity and magnitude of the electric current. Power increase in the theta and decrease in the gamma band was driven by the total electric charge delivered with either increasing the frequency or amplitude of the stimulation current. This inverse effect on the theta and gamma activities was consistently observed in response to different stimulation frequencies and amplitudes. CONCLUSION: Our results suggest a uniform DES effect of amplifying theta and suppressing gamma neural activities in the human brain. SIGNIFICANCE: These findings reveal the utility of simple power-in-band features for understanding and optimizing the effects of electrical stimulation on brain functions.
Subject(s)
Brain , Electroencephalography , Cognition , Electric Stimulation , Head , Humans , Theta RhythmABSTRACT
OBJECTIVE: Conventional selection of pre-ictal EEG epochs for seizure prediction algorithm training data typically assumes a continuous pre-ictal brain state preceding a seizure. This is carried out by defining a fixed duration, pre-ictal time period before seizures from which pre-ictal training data epochs are uniformly sampled. However, stochastic physiological and pathological fluctuations in EEG data characteristics and underlying brain states suggest that pre-ictal state dynamics may be more complex, and selection of pre-ictal training data segments to reflect this could improve algorithm performance. METHODS: We propose a semi-supervised technique to select pre-ictal training data most distinguishable from interictal EEG according to pre-specified data characteristics. The proposed method uses hierarchical clustering to identify optimal pre-ictal data epochs. RESULTS: In this paper we compare the performance of a seizure forecasting algorithm with and without hierarchical clustering of pre-ictal periods in chronic iEEG recordings from six canines with naturally occurring epilepsy. Hierarchical clustering of training data improved results for Time In Warning (TIW) (0.18 vs. 0.23) and False Positive Rate (FPR) (0.5 vs. 0.59) when evaluated across all subjects (p<0.001, n=6). Results were mixed when evaluating TIW, FPR, and Sensitivity for individual dogs. CONCLUSION: Hierarchical clustering is a helpful method for training data selection overall, but should be evaluated on a subject-wise basis. SIGNIFICANCE: The clustering method can be used to optimize results of forecasting towards sensitivity or TIW or FPR, and therefore can be useful for epilepsy management.
ABSTRACT
EEG signal processing is a fundamental method for neurophysiology research and clinical neurology practice. Historically the classification of EEG into physiological, pathological, or artifacts has been performed by expert visual review of the recordings. However, the size of EEG data recordings is rapidly increasing with a trend for higher channel counts, greater sampling frequency, and longer recording duration and complete reliance on visual data review is not sustainable. In this study, we publicly share annotated intracranial EEG data clips from two institutions: Mayo Clinic, MN, USA and St. Anne's University Hospital Brno, Czech Republic. The dataset contains intracranial EEG that are labeled into three groups: physiological activity, pathological/epileptic activity, and artifactual signals. The dataset published here should support and facilitate training of generalized machine learning and digital signal processing methods for intracranial EEG and promote research reproducibility. Along with the data, we also propose a statistical method that is recommended for comparison of candidate classifier performance utilizing out-of-institution/out-of-patient testing.
Subject(s)
Artifacts , Brain , Electrocorticography , Brain/physiology , Brain/physiopathology , Czech Republic , Epilepsy/physiopathology , Humans , Machine Learning , Minnesota , Reproducibility of Results , Signal Processing, Computer-AssistedABSTRACT
OBJECTIVE: When considering all patients with focal drug-resistant epilepsy, as high as 40-50% of patients suffer seizure recurrence after surgery. To achieve seizure freedom without side effects, accurate localization of the epileptogenic tissue is crucial before its resection. We investigate an automated, fast, objective mapping process that uses only interictal data. METHODS: We propose a novel approach based on multiple iEEG features, which are used to train a support vector machine (SVM) model for classification of iEEG electrodes as normal or pathologic using 30 min of inter-ictal recording. RESULTS: The tissue under the iEEG electrodes, classified as epileptogenic, was removed in 17/18 excellent outcome patients and was not entirely resected in 8/10 poor outcome patients. The overall best result was achieved in a subset of 9 excellent outcome patients with the area under the receiver operating curveâ¯=â¯0.95. CONCLUSION: SVM models combining multiple iEEG features show better performance than algorithms using a single iEEG marker. Multiple iEEG and connectivity features in presurgical evaluation could improve epileptogenic tissue localization, which may improve surgical outcome and minimize risk of side effects. SIGNIFICANCE: In this study, promising results were achieved in localization of epileptogenic regions by SVM models that combine multiple features from 30 min of inter-ictal iEEG recordings.
Subject(s)
Electroencephalography/methods , Epilepsies, Partial/diagnosis , Epilepsies, Partial/physiopathology , Adult , Aged , Electrodes, Implanted , Electroencephalography/instrumentation , Female , Humans , Male , Middle Aged , Retrospective Studies , Young AdultABSTRACT
OBJECTIVE: This paper introduces a fully automated, subject-specific deep-learning convolutional neural network (CNN) system for forecasting seizures using ambulatory intracranial EEG (iEEG). The system was tested on a hand-held device (Mayo Epilepsy Assist Device) in a pseudo-prospective mode using iEEG from four canines with naturally occurring epilepsy. APPROACH: The system was trained and tested on 75 seizures collected over 1608 d utilizing a genetic algorithm to optimize forecasting hyper-parameters (prediction horizon (PH), median filter window length, and probability threshold) for each subject-specific seizure forecasting model. The trained CNN models were deployed on a hand-held tablet computer and tested on testing iEEG datasets from four canines. The results from the iEEG testing datasets were compared with Monte Carlo simulations using a Poisson random predictor with equal time in warning to evaluate seizure forecasting performance. MAIN RESULTS: The results show the CNN models forecasted seizures at rates significantly above chance in all four dogs (pâ < 0.01, with mean 0.79 sensitivity and 18% time in warning). The deep learning method presented here surpassed the performance of previously reported methods using computationally expensive features with standard machine learning methods like logistic regression and support vector machine classifiers. SIGNIFICANCE: Our findings principally support the feasibility of deploying trained CNN models on a hand-held computational device (Mayo Epilepsy Assist Device) that analyzes streaming iEEG data for real-time seizure forecasting.
Subject(s)
Deep Learning , Electrocorticography/methods , Electrodes, Implanted , Epilepsy/physiopathology , Seizures/physiopathology , Animals , Dogs , Electrocorticography/instrumentation , Epilepsy/diagnosis , Forecasting , Seizures/diagnosisABSTRACT
Manual and semi-automatic identification of artifacts and unwanted physiological signals in large intracerebral electroencephalographic (iEEG) recordings is time consuming and inaccurate. To date, unsupervised methods to accurately detect iEEG artifacts are not available. This study introduces a novel machine-learning approach for detection of artifacts in iEEG signals in clinically controlled conditions using convolutional neural networks (CNN) and benchmarks the method's performance against expert annotations. The method was trained and tested on data obtained from St Anne's University Hospital (Brno, Czech Republic) and validated on data from Mayo Clinic (Rochester, Minnesota, U.S.A). We show that the proposed technique can be used as a generalized model for iEEG artifact detection. Moreover, a transfer learning process might be used for retraining of the generalized version to form a data-specific model. The generalized model can be efficiently retrained for use with different EEG acquisition systems and noise environments. The generalized and specialized model F1 scores on the testing dataset were 0.81 and 0.96, respectively. The CNN model provides faster, more objective, and more reproducible iEEG artifact detection compared to manual approaches.